Only one column is a true numeric measure, **Amount in USD**. The only other numeric column, **Sr No**, is a row identifier, so its statistics don't describe anything real. These figures come from the profiling query in step 1. **Amount in USD** is stored as text. Only 2,065 of the 3,044 rows convert to a number, so about 979 rows (roughly a third) are blank or non-numeric and are excluded below. | Statistic | Value | |---|---| | Min | 16,000 | | 25th percentile | 470,000 | | Median | 1,700,000 | | Mean | about 18.43 million | | 75th percentile | 8,000,000 | | 95th percentile | 60,000,000 | | Max | 3.9 billion | | Standard deviation (population) | about 121.3 million | - **Heavy right skew:** the mean (about 18.4M) is roughly 10 times the median (1.7M). A few very large deals pull the average up, so the median describes a typical deal better. - **Wide spread:** the standard deviation (about 121M) is roughly 6.6 times the mean. The middle 50% of deals sit between 0.47M and 8M, an interquartile range of about 7.5M. The top 5% of deals are at least 60M, and the maximum of 3.9B is far beyond that. - **Caveat on the standard deviation:** the plain standard deviation query failed with an out-of-range error. The figure above comes from the successful step-1 query, which I labelled `stddev_pop_lt_1e12`. I take that label to mean it excludes absurdly large values (over 1e12), though I didn't re-check its exact filter. Treat it as approximate and lean on the percentiles for spread. **Sr No** runs from 1 to 3,044 with no gaps. Its mean and median are both 1,522.5, and its standard deviation is about 879. That is just what an evenly spaced sequence looks like. The `Date dd/mm/yyyy` column is text, not numeric, so it isn't summarised here.